ARTICLE

https://doi.org/10.1038/s41467-019-09943-y OPEN A fat-tissue sensor couples growth to oxygen availability by remotely controlling insulin secretion

Michael J. Texada 1, Anne F. Jørgensen 1,2, Christian F. Christensen1, Takashi Koyama 1, Alina Malita1, Daniel K. Smith1, Dylan F. M. Marple1, E. Thomas Danielsen 1, Sine K. Petersen1, Jakob L. Hansen2, Kenneth A. Halberg 1 & Kim F. Rewitz 1

Organisms adapt their metabolism and growth to the availability of nutrients and oxygen,

1234567890():,; which are essential for development, yet the mechanisms by which this adaptation occurs are not fully understood. Here we describe an RNAi-based body-size screen in Drosophila to identify such mechanisms. Among the strongest hits is the fibroblast growth factor receptor homolog breathless necessary for proper development of the tracheal airway system. Breathless deficiency results in tissue hypoxia, sensed primarily in this context by the fat tissue through HIF-1a prolyl hydroxylase (Hph). The fat relays its hypoxic status through release of one or more HIF-1a-dependent humoral factors that inhibit insulin secretion from the brain, thereby restricting systemic growth. Independently of HIF-1a, Hph is also required for nutrient-dependent Target-of-rapamycin (Tor) activation. Our findings show that the fat tissue acts as the primary sensor of nutrient and oxygen levels, directing adaptation of organismal metabolism and growth to environmental conditions.

1 Department of Biology, University of Copenhagen, 2100 Copenhagen, Denmark. 2 Cardiovascular Research, Department number 5377, Novo Nordisk A/S, Novo Nordisk Park 1, 2760 Måløv, Denmark. Correspondence and requests for materials should be addressed to K.F.R. (email: [email protected])

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

ulticellular organisms must adapt their metabolism and manipulating Btl-mediated tracheal outgrowth or external O2 Mgrowth to limitations imposed by their environment. levels, or genetically inducing ectopic hypoxia responses via Hph Genetic and physiological pathways such as growth or HIF-1a perturbation, alters the expression of Dilp2, -3, and -5 factor systems regulate these adaptations, and these in turn are in the IPCs and blocks DILP release, leading to reduced systemic influenced by internal and external cues. Among these are insulin signaling. We further report that the primary sensor of nutrient availability, which promotes growth through the insulin/ internal oxygen availability is the fat body. Both hypoxia and insulin-like growth factor (I/IGF) pathways1,2, and oxygen suffi- amino acid (AA) insufficiency block Hph activity in this tissue, ciency, which is also essential for the growth and development and this effect propagates through two divergent pathways of animals3,4. The signaling pathways that regulate these funda- downstream of Hph: HIF-1a-independent inhibition of the mental aspects of development are evolutionarily ancient and Tor pathway alters adipose-tissue physiology, whereas a Tor- well conserved. independent, HIF-1a-dependent pathway leads to the release of I/IGF is the major systemic nutrient-dependent growth reg- one or more humoral factor(s) that strongly inhibit DILP release ulator and acts through conserved pathways to promote cellular and thereby adapt organismal growth to oxygen availability. growth5,6. Many connections between nutritional cues and sys- Thus, Hph/HIF-1a and Hph/Tor pathways in the fat body temic insulinergic regulation have been elucidated in the fruit function as central integrators of both oxygen and AA levels to fly Drosophila melanogaster. Several Drosophila insulin-like pep- adapt growth and metabolism to environmental conditions. tides (DILPs or simply insulin below), primarily DILP2, -3, and Understanding the changes brought about by hypoxia in this -5, are released into circulation from the insulin-producing cells model system may allow greater understanding of human disease (IPCs) of the brain7. Despite their different location, these cells associated with tissue hypoxia such as obesity and diabetes. are functionally homologous with mammalian pancreatic β-cells8. The DILPs signal through a single receptor (InR) to regulate Results both metabolism and growth. DILP expression and release are In-vivo RNAi screen for signals affecting body size.We regulated in part by nutritional information relayed through the undertook a genetic RNAi31,32 screen targeting 1845 genes fat body, an organ analogous to vertebrate adipose and liver tis- encoding the Drosophila secretome and receptome (Fig. 1a, sues with nutrient storage, metabolic, and endocrine functions. Supplementary Data 1). Ubiquitous knockdown using the This tissue secretes insulinotropic (Unpaired-2, functionally daughterless-GAL4 driver (da>) identified 89 genes whose analogous to the mammalian adipokine Leptin9; the peptide fi 10 11 manipulation resulted in signi cant size phenotypes (Fig. 1b, c). hormones CCHamide-2 (CCHa-2) , FIT , and Growth- Among these were several genes associated with body or tissue Blocking Peptides (GPBs) 1 and 212,13; the Activin ligand Daw- 33 14,15 16 growth including Insulin receptor (InR) , Anaplastic lymphoma dle (Daw) ; and the protein Stunted (Sun) ) and insulino- kinase (Alk)34 and jelly belly (jeb)35, Epidermal growth-factor α 17 static (the tumor necrosis factor- homolog Eiger (Egr) ) factors, receptor (Egfr)36, and the prothoracicotropic hormone many of these in response to nutrient-dependent activity of (PTTH) receptor torso37, thus validating our approach. Genes the Target-of-rapamycin (Tor) pathway. Thus, this tissue is a important for ecdysone signaling (knirps, Eip75B, Npc2g, and central nexus for nutritional signals that mediate adaptation to torso37–40) were among the hits that increased pupal body size, nutritional deprivation. consistent with the growth-limiting effects of ecdysone, whereas Organisms require oxygen in addition to nutrients for growth knockdown of the growth-promoting InR produced a strong and development and have therefore developed oxygen-sensing decrease in size. We identified FGF-pathway signaling (Supple- and adaptation mechanisms. In Drosophila and many other mentary Fig. 1) among the strongest hits associated with reduced organisms, hypoxia (low oxygen) restricts systemic growth and 3,18–22 growth. The main hit, the FGFR ortholog btl, and its FGF-like reduces body size , and in humans the limited oxygen ligand Bnl, are critical for proper tracheation of tissues to allow associated with high-altitude living has been linked to slow for gas exchange and oxygen delivery28. growth and developmental delay23,24. These effects occur at oxygen concentrations above those that compromise basic metabolism25,26, indicating that they do not reflect limited FGF signaling affects systemic growth via tracheal effects. aerobic respiration but rather an active adaptation under genetic Ubiquitous btl (da > btl-RNAi)andbnl (da > bnl-RNAi)knockdown strongly reduced growth and pupal size (Fig. 2a, b). Knockdown of control. The conserved transcription factor hypoxia-inducible 41 factor 1 alpha (HIF-1a) is the key regulator required for these the upstream transcription factor trachealess (trh) , also required for tracheal formation, reduced body size to a lesser degree, perhaps adaptive responses. At a rate proportional to oxygen levels, it is fi marked for degradation by HIF-1a prolyl hydroxylase (Hph)27. because of low RNAi ef ciency or Trh-independent signaling later in development42. btl is mainly expressed in the tracheal system but Thus, under hypoxic conditions, HIF-1a is able to perdure and 43 (with its constitutively present beta subunit) induce target-gene it is also active in other cells .Toidentifythetissuecausingthe expression. Although it is well established that nutrients mainly size phenotype, we assayed effects of btl RNAi in the tracheae (btl > affect growth through insulin, the mechanism by which animals btl-RNAi), neurons (elav>), musculature (Mef2>), fat body (ppl>), and glia (repo>). Only tracheal knockdown of btl led to reduced adaptively limit growth under oxygen limitation remains to be fi fi determined. body size (Fig. 2c). RNAi speci city was con rmed with two We describe here an RNA interference (RNAi)-based screen additional independent btl-RNAi lines (Supplementary Fig. 2a). for body-size defects, covering genes encoding potential secreted Knockdown of btl in the tracheae delayed pupariation, thus factors and their receptors, in which we identify breathless (btl)28, prolonging the larval growth period, suggesting that small size encoding a fibroblast growth factor receptor (FGFR) homolog, as resulted from reduced growth rate (Supplementary Fig. 2b). We one of the strongest hits. This receptor is expressed by cells of the therefore measured the growth rate during the third larval instar tracheae, the gas-delivery tubules of the insect respiratory system. (L3) and found indeed that tracheal btl RNAi systemically slowed Its FGF-like ligand Branchless (Bnl) is produced by target tissues body growth (Fig. 2d). Thus, btl is required in the tracheal system in a stereotyped pattern and in response to tissue hypoxia during for normal systemic growth, possibly via effects on oxygen delivery. development29,30 to guide the tracheal terminal branches toward target tissues. When this system is perturbed, internal tissues The btl-RNAi size phenotype is mediated by insulin. Among the experience hypoxia. We find that inducing tissue hypoxia by primary regulators of systemic growth in Drosophila are the

2 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE

a c 30 Increased size Increased size RNAi knockdown (40 genes) Secretome/ 20 Receptome (1845 genes)

Size (%) 10 Δ Reduced size (49 genes) Z score cut-off ± 2 0 vg tor jeb kni chb shg ng3

(1756 genes) Hml Egfr kek1 Hnf4 spz3 Syx8 IdICp Rab7 Wsck Smox Ag5r2 Gr98c Or65c Gr39a Or56a Npc2g Eip75B Cpr31A TM4SF Snap29 Spn53F AP-2mu CG5758 CG3088 CG5863 CG2145 Dh44-R1 Muc12Ea CG32036 CG10186 CG13056 CG10587 unc-13-4A

b 6

0 Cht7 btl CG33128 CG8997 rig Galphas siz Alk InR CG32395 Oct-TyrR CG33462 Trap1 Gr47b Mipp2 Muc96D CG8083 Npc2b GNBP-like3 Yp1 polyph Yp2 CG13641 CG32066 CG13589 eca dor CG14406 CG31681 IM33 CG11912 cm babo CG11377 Lcp65Ag1 CG8343 CG3009 CG17278 CG11362 Gr98d ana Jon44E PPO2 Iectin-24A CheA75a CG4151 GV1 CG13049 CG6592

4 –10 –20 (40 genes) Increased size –30

2 Size (%) Δ –40 Reduced size

–50 0

d Torso (Tor) EGFR Size (Z score) Δ –2 Reduced size (49 genes) Growth –4

RNAi targeted genes InR (Insulin) Alk Btl (FGFR) –6

Fig. 1 In-vivo RNAi screen for factors regulating body size. a RNAi screen of 1845 genes of the Drosophila secretome and receptome selected by gene ontology analysis. Upstream activating sequence (UAS)-inducible RNAi constructs against these genes were expressed ubiquitously using da-GAL4, and pupal sizes were determined (data are presented in Supplementary Data 1). b, c Distribution of the pupal sizes from the screen (b) revealed 89 gene hits that resulted in a significant increase (c: blue, top) or decrease (c: orange, bottom) in body size (a Z-score > ± 2 was used as a cutoff) by comparison with the mean of all lines. n = 9–55. d Genes and pathways identified in the screen having known size-regulatory function, as well as breathless (btl), are marked with arrows. Error bars indicate SEM. Underlying data are provided in Supplementary Data 1

DILPs7. To assess whether the body-size phenotype of btl observed size phenotype, we asked whether rescuing circulating knockdown is mediated by altered insulin signaling, we first insulin levels by overexpressing Dilp2 would also rescue growth. examined the effects of directly reducing insulin signaling. As Ubiquitous btl knockdown driven by armadillo-GAL4 (arm>) led expected, globally reducing insulin signaling (da > InR-RNAi)or to reduced body size that was rescued by Dilp2 overexpression blocking Tor-dependent fat-body nutrient signaling through (Fig. 3g), supporting that the size phenotype is caused by a expression of Tsc1 and Tsc2 (ppl > Tsc1/2) strongly reduced pupal systemic reduction in insulin signaling downstream of Btl. size and prolonged larval development (Fig. 3a, Supplementary Alternatively, the observed reduced body size might involve Fig. 3a), phenocopying tracheal btl loss. To determine whether increased ecdysone signaling, which antagonizes insulin signaling btl > btl-RNAi might cause insulin expression changes, we quan- and suppresses larval body growth46. However, btl knockdown tified Dilp2, -3, and -5 transcripts in feeding L3s, ~ 24 h before led to reduced expression of the ecdysone-induced gene E75B, pupariation. Indeed, very strong reduction (>90%) of Dilp3 which indicates reduced ecdysone-induced growth restriction expression was observed in RNAi animals, accompanied by a (Supplementary Fig. 3b). Together, these results suggest that the smaller decrease in Dilp5 transcripts (Fig. 3b). reduced body size observed with btl knockdown is mediated by Insulin activity is also regulated at the level of DILP release44, effects in the tracheal system that lead to altered expression and and we therefore investigated whether tracheal btl loss might reduced secretion of DILP2, -3, and -5 from the IPCs, resulting in induce IPC retention of DILP2, -3, and -5. Knockdown induced reduced insulin-driven systemic growth. an accumulation of these peptides in the IPCs (Fig. 3c), despite the reduced expression of Dilp3 and Dilp5. This suggests that btl > btl-RNAi induces a strong decrease in DILP secretion, which btl RNAi leads to hypoxia via reduced tracheation. Hypoxic should reduce peripheral activity downstream of InR. Insulin tissues secrete the FGF-like ligand Bnl, inducing tracheal growth signaling leads to the phosphorylation of the kinase Akt and to toward these tissues via its receptor Btl. To assess whether btl suppression of InR expression as part of a sensitivity-regulating knockdown reduced tracheation of internal tissues, we observed feedback circuit45. We indeed observed InR upregulation and tracheal branches associated with the fat body in larval abdominal phosphorylated Akt (pAkt) reduction in btl > btl-RNAi animals segments 2 and 3 (Fig. 4a), quantifying the number of branches (Fig. 3d–f), indicating decreased insulin signaling, presumably and total tracheal length47. Control tracheae exhibited extensive arising from blocked IPC DILP release. To determine whether the terminal branching, which was significantly reduced in btl > btl- observed reduction in insulin signaling was the direct cause of the RNAi animals, along with the total length of tracheal processes

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

ab10

–10 Size (%) Δ –30 *** >+ da btl-RNAi bnl-RNAi trh-RNAi > > > –50 *** *** da da da

btl-RNAibnl-RNAitrh-RNAi > > > da da da

c d

10 Control Global Tracheae Neurons Muscle Fat body Glia

0 2.0 btl>+ btl-RNAi/+

–10 1.5 btl>btl-RNAi

Size (%) –20 1.0 Δ

0.5 –30 Weight (mg/larva) *** *** 0 –40 + 75 80 85 90 95 Time (hours after egg lay) btl-RNAibtl-RNAibtl-RNAibtl-RNAibtl-RNAibtl-RNAi btl-RNAi/> > > > > > da btl ppl elav Mef2 repo

Fig. 2 Tracheal loss of breathless reduces systemic growth and body size. a, b Quantification of pupal sizes and representative images (b) of animals with ubiquitous knockdown of breathless (btl)-pathway components compared with the controls (da > +; da > crossed to wild type, w1118). Values are percent change in pupal size vs. control. a: n = 31–64. c Knockdown of btl in whole animals (da > btl-RNAi) or tracheae alone (btl>), but not in neurons (elav>), muscle (Mef2>), fat-body tissue (ppl>), or glia (repo>) leads to small animals compared with da > + controls. Shown also is the RNAi control (btl-RNAi/+ ; btl-RNAi crossed to w1118). Values are percent change in pupal size vs. the btl > + controls (btl > crossed to w1118). n = 33–86. d Trachea-specific knockdown of btl reduces larval growth rates compared with driver (btl > + ) and RNAi controls (btl-RNAi/+ ). Statistics: one-way ANOVA with Dunnett’s multiple- comparisons test. ***P < 0.001, compared with the control. Error bars indicate SEM. Underlying data are provided in the Source Data file

(Fig. 4b, c), implying a reduced oxygen supply. In btl > btl-RNAi Thus, our findings place insulin downstream of the growth of the animals, we observed strong upregulation of bnl (Fig. 4d), which tracheal system and oxygen levels, mediating their effects on is induced in response to hypoxia29, indicating that internal tis- systemic growth. sues were indeed experiencing hypoxia. We rationalized therefore that these animals would be sensitized to low-oxygen conditions. To test this, we reared control and btl > btl-RNAi animals under Hypoxia pathway activation mimics loss of btl. Thus, ubiqui- hypoxic conditions (5% O2 vs. the normal 21% atmospheric level) tous or trachea-specific loss of btl induces internal tissue hypoxia, and measured their survival to the pupal stage. Most control which via reduced insulin signaling leads to reduced organismal animals survived this environment, reflecting their ability to growth. To confirm that hypoxia per se is the cause of this adapt to reduced oxygen levels, whereas none of the btl-RNAi reduction and to assess whether this inhibition is part of an animals survived to pupariation (Fig. 4e), presumably dying from adaptive response rather than a pathological result of low oxygen hypoxia-induced damage. Taken together, our data suggest that levels, we investigated the effects of external hypoxia on larval 1118 the insulinergic changes described above operate downstream of growth. Wild-type (w ) animals reared under 5% O2 displayed tissue hypoxia to adapt growth to oxygen availability. reduced pupal size, despite a delay in pupariation that extended To investigate whether insulinergic changes might regulate the the larval growth period, phenocopying btl > btl-RNAi and indi- tracheae themselves, we assessed the effects of altering insulin cating strong growth reduction (Fig. 5a, b, Supplementary signaling in these tissues. Although overexpression of Dilp2 Fig. 4a). rescued growth of arm > btl-RNAi animals (Fig. 3g), it did not The signaling pathway that transduces oxygen levels to induce restore the growth of their tracheae (Supplementary Fig. 3c). cellular adaptation is conserved between Drosophila and mam- Furthermore, tracheal knockdown of InR or Akt had no effect on mals. The main effector protein of this pathway is the systemic growth or pupal size, whereas loss of Tor signaling in the transcription factor HIF-1a, encoded in Drosophila by the gene tracheae led to a strong reduction in size (Supplementary Fig. 3d). similar (sima)48. Under normoxic conditions, this protein is

4 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE

ab c *** 10 500 6 *** btl>+ btl>+

400 ** btl>btl-RNAi btl>btl-RNAi –10 4 300 –30 ***

Size (%) 200 Δ *** *** 2

–50 Relative expression 100 ** Normalized fluorescence ** –70 *** 0 0 Dilp2 Dilp3 Dilp5

btl-RNA Tsc1/2 > InR-RNA> > btl ppl da α-DILP2 α-DILP3 α-DILP5

debtl>+ f g 600 btl>btl-RNAi 1.5 btl>+ 10 pAkt btl>btl-RNAi * 0 400 Akt 1.0

–10

Tubulin Size (%) pAkt ratio ** Δ 200 0.5 Relative expression ** –20 >+ btl btl-RNAi *** > 0 btl 0 –30 InR pAkt/Tub pAkt/Akt Dilp2 > btl-RNAi > arm arm btl-RNAi, Dilp2 > arm

Fig. 3 Breathless modulates systemic growth by altering DILP signaling. a Trachea-specific knockdown of breathless (btl) mimics loss of insulin signaling induced by ubiquitous InR knockdown (da > InR-RNAi) and fat-body inhibition of Tor (ppl > Tsc1/2), which remotely inhibits release of insulin-like peptides (DILPs) from the insulin-producing cells (IPCs). Values are percent change in pupal size vs. the btl > + controls (btl> crossed to w1118). n = 15–79. b, c Levels of whole-animal Dilp transcripts and DILP peptides in the IPCs in btl > btl-RNAi animals with knockdown of btl in the trachea compared with btl > + controls. Representative images of DILP2, -3, and -5 immunostainings are shown below. b: n = 5. c: n = 44–45. d Increased expression of InR in btl > btl- RNAi animals compared with btl > + controls suggests reduced systemic insulin signaling. n = 5. e, f Immunoblotting showing whole-animal levels of phosphorylated Akt (pAkt) (e)inbtl > btl-RNAi animals compared with btl > + controls, quantified in f from three independent replicates normalized to Tubulin (Tub) or Akt levels. f: n = 3. g Ectopic expression of Dilp2 rescues the btl-knockdown size phenotype. Ubiquitous weak knockdown of btl using armadillo-GAL4 (arm > btl-RNAi) leads to small pupae (left). Co-expressing Dilp2 rescues this size phenotype (right), while having no significant effect on control animals (middle). Values are percent change in pupal size versus arm > + controls. n = 14–113. Statistics: one-way ANOVA with Dunnett’s test for multiple comparisons and Student’s t-test for pairwise comparisons. *P < 0.05, **P < 0.01, ***P < 0.001, compared with the control. Error bars indicate SEM. Underlying data are provided in the Source Data file marked for proteasomal degradation through the oxygen- well, as Dilp3 expression is greatly reduced by up to 90%. To dependent activity of Hph, also called Fatiga (Fga) in confirm IPC DILP retention, we expressed epitope-tagged Drosophila49,50. To differentiate between pathological damage DILP251 in these cells using Dilp2-GAL4 (Dilp2>) and measured vs. a genetic adaptive response, we activated the hypoxia- levels of circulating tagged insulin by enzyme-linked immuno- adaptation program by mutation of Hph, thus de-repressing sorbent assay (ELISA). Starvation conditions, known to reduce Sima under normoxic conditions. Similar to hypoxia or btl hemolymph DILP2 levels52, greatly reduced tagged DILP2 titer knockdown, loss of Hph function resulted in growth inhibition (Fig. 5e). However, feeding animals moved to hypoxia showed an and developmental delay despite the animals’ normoxic environ- even stronger decrease in insulin release. Thus, hypoxia indeed ment (Fig. 5a, b, Supplementary Fig. 4a), confirming the genetic blocks DILP release from the IPCs. Consistent with decreased control of the growth phenotype. systemic insulin signaling, both the low-O2 environment and Hph To evaluate whether the insulin-related changes seen in btl- loss led to reduced pAkt levels (Fig. 5f, Supplementary Fig. 4b). knockdown animals are part of a genetic hypoxia-adaptation We further investigated this using an in-vivo insulin-pathway program, we measured Dilp expression and DILP retention reporter that reflects intracellular insulin-pathway activity based within the IPCs in hypoxic wild-type animals and in normoxic on plasma-membrane localization of green fluorescent protein Hph mutants. Dilp3 and Dilp5 transcription was strongly reduced (GFP)53. Under normoxia GFP was localized to the membrane of in both manipulations (Fig. 5c). Furthermore, both conditions led fat-body cells, indicating active insulin signaling, whereas under to increased IPC DILP2 and DILP5 retention (Fig. 5d). The hypoxia, the reporter remained cytoplasmic, confirming reduced observed DILP3 level may represent retention of this protein as signaling (Fig. 5g).

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

a b btl>+ btl>btl-RNAi GFP GFP T1 T2 GFP

T3

A1 A2 A3 A4 A5 PB PB A6 SB SB A7

A8/A9 TB

cdbtl>+ e 1.5 2000 100 btl>btl-RNAi 5% O2 80 1500 * 1.0 60

1000 40 0.5 Relative expression 500

Normalized trachea length 20 *** Animals surviving to pupa (%)

0.0 0 0 *** bnl >+ btl btl btl-RNAi btl-RNAi > > btl btl

Fig. 4 Airway branching is reduced in breathless-knockdown animals. a Tracheae ramifying to the fat body were measured in larval segments A2/A3, illustrated in overview here. b, c Knockdown of breathless (btl) in tracheae leads to a reduction in branching of the terminal tracheal cells of the fat body branch compared with the btl > + control (btl > crossed to w1118). Representative images (b) and quantification (c) of the fat-body branch of larval tracheal terminal cells are shown. Scale bar, 50 µm. PB, primary branch; SB, secondary branch; TB, tertiary branch. c: n = 4. d Whole-body transcript levels of branchless (bnl), encoding the hypoxia-inducible FGF ligand of Btl, is strongly upregulated in larvae with trachea-specific btl knockdown, suggesting that tissues are experiencing hypoxia in these animals. n = 4–5. e Loss of btl renders animals unable to survive in a low-oxygen environment (5% O2 levels), whereas most btl > + controls are able to develop to pupariation. n = 3. Statistics: Student’s t-test for pairwise comparisons. *P < 0.05, ***P < 0.001, compared with the control. Error bars indicate SEM. Underlying data are provided in the Source Data file

Similar to animals with btl knockdown, hypoxic wild-types and elevation, indicating that this tissue experiences hypoxia with normoxic Hph mutants exhibited reduced ecdysone signaling tracheal insufficiency (Fig. 6a). To confirm this hypoxic state, we (measured as expression of E75B) and upregulation of bnl made use of a genetic hypoxia indicator based on GFP fused with (Supplementary Fig. 4c, d). To rule out ecdysone deficiency as an the oxygen-dependent degradation domain (ODD) of HIF-1a54. underlying cause of the hypoxic growth effect, we supplemented Under normoxic conditions, this GFP-based reporter is degraded wild-type animals with ecdysone by feeding and found that this through Hph activity, whereas red fluorescent protein (RFP) treatment even further decreased pupal sizes under hypoxic expressed from the same regulatory sequences acts as a ratio- conditions (Supplementary Fig. 4e), demonstrating that reduced metric control. In the fat body, the GFP::ODD reporter was ecdysone does not underlie hypoxia-induced growth restriction. rapidly degraded under normoxia, leading to a low GFP:RFP ratio Taken together, these findings are consistent with an underlying (Fig. 6b, Supplementary Fig. 5a). In contrast, an increased ratio phenomenon common to btl knockdown, hypoxic wild-types, was observed in animals reared under 5% O2, indicating reduced and normoxic Hph mutants—namely, the activation of the oxygen-dependent degradation of the ODD-linked GFP. Toge- genetic hypoxia-adaptation program that limits growth by ther, these data show that the fat body experiences physiologically reducing insulin signaling. significant oxygen stress under our manipulations.

Loss of btl or low O2 cause adipose-tissue hypoxia. The fat body An adipose-derived hypoxia signal represses insulin release.We is reported to secrete factors that modulate IPC insulin expression next asked whether the brain and the IPCs alone could sense and release9–11,14–17. To investigate whether this tissue is also hypoxia and induce changes in insulin physiology, or whether the involved in hypoxia adaptation, we first measured fat-body bnl fat body might be necessary for organismal growth adaptation to transcription under tracheal btl knockdown and observed strong hypoxia. To investigate this, we cultured larval brains alone or with

6 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE

ab e 20 1000 *** 0 *** 100 l)

–20 μ

Size (%) *** (pg/ Δ 10 –40 Hemolymph DILP2HF

*** *** –60 ) ) ) 1 ) ) 2 2 2 2 2 Oxygen 21% 21% 5% (5% O (5% O (21% O (21% O 9 (21% O Food +–+ 9 1118  1118 1118 w w w Hph Hph c f 200 w1118 (21% O ) 2 pAkt 1118 w (5% O2)  Hph 9 (21% O ) 150 2 Akt

100 ** Tubulin

** ) ) ) 2 2 2 *** Relative expression 50 (5% O (21% O 9 (21% O 1118  1118 *** w w Hph 0 Dilp2 Dilp3 Dilp5 g 1118 1118 w (21% O2) w (5% O2) d *** 10 1118 w (21% O2)

1118 w (5% O2)

9 7.5 Hph (21% O2) *

5 *** *** *** tGPH tGPH

2.5 Normalized fluorescence

0

tGPH tGPH α-DILP2 α-DILP3 α-DILP5 DAPI DAPI

Fig. 5 Hypoxia and induction of hypoxia adaptation block growth and DILP secretion. a, b Quantification of pupal size (a) and representative images (b)of 1118 animals exposed to hypoxia (5% O2) or mutation of HIF-1a prolyl hydroxylase (Hph) compared with controls (w in normoxia). Values are percent change in pupal size vs. w1118-in-normoxia controls. a: n = 24–34. c, d Transcript levels of Dilp genes in whole animals (c) and DILP peptide levels (d) in the insulin-producing cells (IPCs) of hypoxic wild-types and normoxic Hph mutants compared with normoxic wild types. Representative images of IPC DILP2,

-3, and -5 immunostaining are shown below. c: n = 6. d: n = 29–42. e Hemolymph HA::DILP2::FLAG (DILP2HF) levels in normoxic (21% O2, on normal food), hypoxic (5% O2; normal food), and starved (21% O2; 1% agar) animals determined by ELISA. n = 5. f Immunoblotting shows that levels of phosphorylated Akt (pAkt) is reduced under hypoxia and in Hph mutants compared with btl > + controls when normalized to alpha-Tubulin (Tub) or total Akt levels. g tGPH reporter of insulin signaling confirms that hypoxia reduces systemic insulin signaling in peripheral tissues, represented here by the fat body. Under conditions of normal insulin activity (e.g., fed normoxia, left), the tGPH sensor is localized to the plasma membrane, whereas under conditions of low signaling (e.g., fed hypoxia, right) the sensor becomes primarily cytoplasmic. Scale bar, 20 µm. Statistics: one-way ANOVA with Dunnett’s multiple- comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, compared with the control. Error bars indicate SEM. Underlying data are provided in the Source Data file fat bodies for 16 h ex vivo under normoxia or hypoxia and assessed fat bodies, however, did exhibit reduced Dilp expression under the insulin-related effects of these manipulations (Fig. 6c). Whereas hypoxia compared with normoxic cultures. These observations hypoxiclarvaeexhibitreducedDilp3 and -5 expression (Fig. 5c), we demonstrate that the brain alone is insufficient for hypoxia adap- observed no decrease in insulin-gene expression in brains cultured tation, and that a humoral signal(s) from the fat body regulates alone under hypoxia vs. normoxia (Fig. 6d). Brains co-cultured with Dilp3 and -5 expression as part of this response.

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

abbtl>+ c btl>btl-RNAi 600 ** 3 **

400 2

200 1 Ratio (GFP/RFP)

Relative expression Brain Fat body

0 0 Schneider's medium bnl 21% 5%

O2 de* f *** *** n.s. 5 400 Dilp2 Dilp3 Dilp5 4 *** 4 300 3 3 200 ** ** * 2 2

100 1 Relative expression 1 Normalized fluorescence Normalized fluorescence

0 0 0 21% O2 +– +– +– +– +– +–

5% O2 – + –+ –+ –+ –+ –+ w1118 fat body – – + + – – + + – – + + α-DILP2 α-DILP2

21% O2 + – + – + + g 10 5% O2 –++ –+ – – w1118 fat body + + – – + – Hph9 fat body –––+– – 0

h ppl>Cas9/+ i ppl>Cas9/+ –10 KO KO 200 ppl>Cas9, Hph ppl>Cas9, Hph 3 *** Size (%)

Δ –20 150 *** 2 –30 100 *** –40 1 50 /+ Relative expression KO KO KO * h p H Normalized fluorescence 0 ppl>Cas9/+Cg>Cas9/+ 0 UAS-Hph Dilp3 ppl>Cas9,Cg>Cas9, Hph

α-DILP2

Fig. 6 A fat-body hypoxia sensor represses insulin secretion from the brain. a Transcriptional upregulation of branchless (bnl) in the fat body of animals with trachea-specific btl knockdown indicates hypoxia. n = 5–6. b A transgenic reporter of hypoxia indicates that the fat body experiences low oxygen levels that inhibit Hph when animals are incubated under 5% O2 compared with 21% O2 conditions. Ratio of integrated GFP (carrying the oxygen-dependent degradation domain of HIF-1a) and RFP (without this domain) signals in fat body. n = 10. c Illustration of ex-vivo organ culture technique. d Transcript levels of Dilp genes in brains alone or brains and fat bodies cultured ex-vivo under hypoxia, compared with normoxic cultures. Dilp transcription was unchanged or increased (nonsignificant) when brains alone were exposed to hypoxia, whereas transcription of all three genes was reduced in brains co-incubated with fat-body tissue, closely mimicking the changes seen in whole animals. n = 5–6. e Ex-vivo co-culture of wild-type brains with wild-type fat bodies causes DILP2 retention under hypoxia. Brains incubated alone without fat body are unable to induce DILP2 retention under hypoxia. Thus, the presence of the hypoxic fat body actively represses DILP secretion. n = 50–58. f Co-culture of wild-type brains with Hph-mutant fat bodies induces DILP2 retention under normoxia (right). Representative images of DILP2 immunostainings in the IPCs are shown below. n = 40–67. g–i CRISPR/Cas9-induced fat-body-specific Hph knockout (HphKO) reduced pupal size (g) and transcript levels of Dilp3 (h), and caused DILP2 retention in the IPCs (i). Representative images of DILP2 immunostainings in the IPCs are shown below. The ppl> and Cg> fat-body GAL4 drivers were used to express UAS-driven Cas9 and gRNAs designed to delete part of the Hph locus specifically in the fat body. Controls: ppl > Cas9/+, Cg > Cas9/+, and UAS-HphKO/ + (HphKO/+). g: n = 23–44. h: n = 5–6. i: n = 11-16. Statistics: Student’s t-test for pairwise comparisons and one-way ANOVA with Dunnett’s for multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, compared with the control. Error bars indicate SEM. Underlying data are provided in the Source Data file

8 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE

To confirm the requirement of fat-body genetic hypoxia be altered by hypoxia, and, if so, how AA levels and Hph and responses for DILP2 retention, we cultured wild-type brains Sima/HIF-1a activity affected this alteration. We reared wild-type either alone or with wild-type or Hph-mutant fat bodies, under animals on standard 1× food in normoxia and hypoxia, and normoxia and hypoxia. Co-culturing wild-type brains and fat measured levels of phosphorylated ribosomal protein S6 (pS6), a bodies under normoxia induced a drop in DILP2 retention downstream effector and proxy for Tor signaling57, in the fat compared with brains alone, consistent with previous studies bodies of feeding late-third-instar larvae. Hypoxic rearing very showing that the fat body releases humoral signals that promote strongly reduced pS6 levels (Supplementary Fig. 5b), indicating insulin secretion in well-fed animals reared under normal oxygen that constitutive hypoxia induces a large reduction in Tor activity. conditions9,12,16,55. Brains cultured alone did not exhibit DILP2 We next measured levels of pS6 in wild-type animals and Hph changes between hypoxia and normoxia; however, when brains and sima mutants, under normoxia or 16 h hypoxic conditions, were cultured with wild-type fat-body tissue, hypoxia strongly with and without 16 h AA starvation. Wild-type animals kept induced DILP2 retention (Fig. 6e). These changes indicate that on standard 1× food under normoxia exhibited high levels of hypoxia-induced inhibitory fat-body factor(s) are required for pS6, indicating strong Tor activity (Fig. 7b), and 16 h hypoxia insulin retention. To assess the genetic control of this signal, we treatment led to a decrease in pS6 staining, as did AA starvation, co-cultured wild-type brains with Hph-mutant fat bodies under which is known to inhibit Tor. Combining these treatments, normoxia (Fig. 6f) and found again that DILP2 was retained in hypoxia and AA starvation, did not further reduce fat body pS6 the IPCs. levels beyond that of starvation alone, suggesting either that they Next, as this fat-body oxygen sensor regulates insulin act through the same pathway or that Tor is already maximally physiology, we wished to examine its necessity in the regulation inhibited by protein starvation. of systemic growth. To ensure efficient gene-activity disruption, Hph-mutant animals exhibited low fat-body pS6 staining that we made UAS-inducible CRISPR/Cas9 constructs to delete was not affected by oxygen or AA levels. This is consistent with Hph specifically in the fat tissue. Under normoxia, pupal size the hypoxic response as well as the lack of growth response was strongly reduced by fat-body-specific disruption of Hph, of Hph mutants to dietary AAs, indicating that Hph activity driven by either of two independent fat-body-GAL4 lines (Fig. 6g), is required for Tor-pathway function. A known feedback loop similar to the size reduction observed with hypoxia or traditional connects HIF-1a/Sima to Tor inhibition58,59, so we next Hph mutation. Furthermore, we found that deletion of Hph in investigated whether Sima is required for hypoxia-induced Tor the fat body led to reduced Dilp3 transcription and to DILP2 inhibition. In contrast to loss of Hph, animals lacking sima still retention in the brain (Fig. 6h, i) under normoxia. Fat-body- exhibited high levels of Tor activity under standard food and specific Hph knockout therefore recapitulates major components normoxic conditions, indicating that Sima is not required for of the organismal hypoxic response. Taken together, these results Tor activity in these conditions. Furthermore, these animals also indicate that the activity of the Hph-regulated hypoxia-adaptation retained Tor inhibition under restrictive conditions of oxygen or program within the fat body is the determinant, via secreted AA deficiency (or both), indicating that Sima is not required for inhibitory factor(s), of IPC insulin retention and thereby of Tor inhibition in these situations. Therefore, oxygen-regulated systemic growth rate. Hph activity, but not Sima activity, is required for Tor activity and response to AAs. To assess the physiological significance of Tor activity in the fat, Hph is required for Tor activity, independently of HIF-1a. we measured the size of lipid droplets (LDs), a parameter thought Beyond its role in coupling oxygen to systemic growth demon- to reflect Tor activity55, under the same environmental conditions strated above, the fat body also links intake of dietary AAs to as above. LDs of wild types were relatively small and their sizes systemic growth through a Tor-dependent nutrient-sensing increased under hypoxia or AA starvation (Fig. 7c), consistent mechanism that regulates release of insulin-governing factors. with the reduced Tor activity observed under these conditions We therefore next explored the possible relationships between (Fig. 7a). On the other hand, loss of Hph led to greatly enlarged oxygen and AA sensing in growth regulation. We reared wild- LDs in the fat body and their size was not further altered by type animals and Hph and sima mutants56 on foods containing hypoxia or AA starvation, consistent with the data above that 0.1×, 0.3×, or 1× (standard food) concentrations of yeast (the suggest that Hph is required for transducing these signals through customary dietary protein source) under normoxia or 5% O2 and a Tor-dependent mechanism. Both hypoxia and AA starvation assessed the animals’ growth responsiveness to dietary protein. still led to increased LD size in sima mutants, indicating that Whereas wild-type animals displayed a robust scaling of body size Sima activity is not required for these effects, consistent with to dietary protein levels under normoxia, hypoxia attenuated the finding that these animals also inhibit Tor appropriately in or blocked this response (Fig. 7a). Similarly, inducing genetic response to AA starvation and hypoxia. Thus, the Hph-dependent hypoxic responses by Hph mutation rendered animals growth- and HIF-1a/Sima-independent Tor-activity differences observed unresponsive to protein even under normoxia, indicating that under varied environmental and genetic perturbations lead to normal growth responses to dietary AAs require Hph signaling. physiological changes in fat distribution within the fat body. To address whether this effect requires HIF-1a, we analyzed the As our data indicate that Tor-dependent AA responses require growth response of HIF-1a/sima mutants to dietary protein. Hph activity, we next tested directly whether Hph might be Under normoxic conditions, sima-mutant animals responded involved in sensing AA levels, using the GFP::ODD Hph-activity to dietary AAs to the same degree as wild types. Thus, unlike reporter construct, and observed that 4 h protein starvation led Hph, Sima is not required for AA-dependent growth regulation. to a significant increase in fat-body GFP::ODD perdurance However, in contrast to wild-type animals, sima mutants still (Fig. 7d). As Hph activity targets ODD-containing proteins for responded to dietary protein even under hypoxic conditions, and destruction, this result suggests that Hph activity is reduced by their hypoxia-induced size reduction was significantly rescued insufficiency of AAs as well as of oxygen. Thus, both AAs and (Fig. 7a), indicating that Sima is required for hypoxia-induced oxygen may induce growth through modulating Hph activity, growth restriction. which is required for growth. As a number of known growth-regulatory factors released by As our data suggest that both AAs and oxygen affect Tor the fat body are regulated by Tor activity in response to AA through Hph, and Tor in the fat body is known to regulate the availability10,11,16,17, we investigated whether Tor signaling might release of factors that control DILP secretion, we wished to

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

a *** 3500 sima-/- rescue 3000

2500

2000 Pupal size

1500

1000 Yeast 0.1× 0.3× 1× 0.1× 0.3× 1× 0.1× 0.3× 1× 0.1× 0.3× 1× 0.1× 0.3× 1×

21% O2 5% O2 21% O2 21% O2 5% O2 Wild type (w1118) Hph9 sima–/–

b *** ***

300

*** 200 ***

100 ns Fat-body pS6 levels

0

Hypoxia (5% O2) –+– + –+– +–+– + AA starvation ––++ ––++ ––++

Wild type (w1118) Hph–/– sima–/–

c d *** *** 21% O ns 2 20 *** ns ** 2.5 ** * ns ns * ) 2 2.0 m 15 μ 1.5 10 1.0 Ratio (GFP/YFP) 5 0.5 Lipid droplet size (

0 0.0 AA – + starvation

Hypoxia (5% O2) –+––+––+– AA starvation ––+––+––+

Wild type (w1118) Hph–/– sima–/–

Fig. 7 Hph-dependent amino-acid and oxygen sensing via Tor is HIF-1a/Sima-independent. a Growth response to dietary amino-acid levels (0.1×, 0.3×, and 1118 standard 1× dietary yeast concentrations) of wild types (w ) and Hph and sima mutants under normoxia (21% O2) and hypoxia (5% O2). Note that sima mutants are significantly larger than wild types on 1× yeast food in hypoxia, whereas their sizes are not significantly different in normoxia (p = 0.26), further indicating that Sima is required for hypoxia-induced growth suppression. n = 8–71. b Fat-body staining against phosphorylated ribosomal protein S6 (anti-pS6), a downstream indicator of Tor activity, indicates that the Tor pathway is suppressed by 16 h amino acid starvation (1× sugar-only medium), hypoxia (5% O2), or Hph mutation. sima mutants exhibit Tor-pathway activation under normoxia, and thus Sima is not required for Tor activity. Tor activity decreases under hypoxia and with amino-acid deprivation, indicating that Sima activity is also not required for Tor suppression. Combined treatments of amino-acid starvation and hypoxia do not further suppress Tor. Representative images of pS6 stain are shown. n = 8–30. c Sizes of fat-body lipid droplets 1118 examined by CARS microscopy in wild types (w ) and Hph and sima mutants with 16 h hypoxia (5% O2) or amino acid starvation (1× sugar-only medium). Lipid-droplet size increases under hypoxia and protein starvation in wild types and sima mutants, but not Hph mutants. Representative CARS images are shown. n = 10–18. d Hph activity reflects amino acid availability. Fat-body imaging of the GFP::ODD Hph-activity reporter indicates that GFP::ODD degradation is reduced under 4 h amino acid starvation (1x sugar food), suggesting that Hph activity is blocked by lack of amino acids. n = 20. Statistics: Student’s t-test for pairwise comparisons and one-way ANOVA with Dunnett’s for multiple-comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, compared with the control. Error bars indicate SEM. Underlying data are provided in the Source Data file

10 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE determine whether Tor activity per se might contribute to factors, to rule out their involvement. None of these factors observed hypoxia-induced growth effects. We therefore stimu- appears to affect hypoxia-induced growth restriction. Thus, the lated the Tor pathway in the fat body by expressing the activator factor released by the fat body appears to be a previously Rheb (ppl > Rheb). Although this manipulation dramatically uncharacterized Tor-independent signal that is released down- increased Tor activity under hypoxia, it did not rescue the size stream of Hph and HIF-1a/Sima as an inhibitor of insulin phenotype (Supplementary Fig. 6a, b), indicating that the size- secretion. regulatory effects of hypoxia do not arise from alteration of Tor activity in the fat body. To further confirm the Tor independence of hypoxic growth reduction, we assessed the effects of individual Hypoxia reduces systemic growth via HIF-1a in the fat body. known Tor-regulated growth modulators. Of the known fat- As the data above suggested that HIF-1a/Sima is required in the body-derived factors, Egr is the only one that is reported to be fat body for hypoxia-adaptation effects on body size, we investi- insulinostatic, although its specific effects do not exactly match gated the effects of fat-body-specific manipulations of HIF-1a/ those we observed under hypoxia17. We expressed RNAi against Sima function. First, we examined whether fat-body-specific egr in the fat body or against the gene encoding its receptor, RNAi against sima could block the DILP retention induced by Grindelwald (Grnd), in the IPCs. Consistent with the idea of a hypoxia. We observed a small but significant reduction in IPC Tor-independent mechanism, we found that neither egr nor grnd DILP2 levels and partial rescue of the hypoxia-induced growth RNAi rescued the low-oxygen body-size reduction, suggesting inhibition with this manipulation (Supplementary Fig. 6c, d), that they do not play a role in hypoxia-induced growth restriction which might indicate weak knockdown. Therefore, we repeated (Supplementary Fig. 6b). We also assessed other fat-body-derived this experiment using tissue-specific CRISPR/Cas9-mediated factors and their receptors such as CCHa-2, Sun, GBPs 1 and 2, disruption of sima and observed a much stronger reduction in and Daw, although these are reported to be growth-promoting IPC DILP2 levels under hypoxia (Fig. 8a). Furthermore, Dilp3

abppl>Cas9/+ ppl>Cas9/+ c ppl>Cas9, simaKO ppl>Cas9, simaKO 2 400 40 5% O 5% O 2 2 5% O2

300 * 30 1 *** *** 200 20 Size (%) Δ

Normalized fluorescence 0 100 10 Relative expression

0 0 α + + -DILP2 Dilp3 / KO KO Cas9/ > sima ppl AS-sima U Cas9, > pl p d Systemic growth

Secreted Tracheation factor(s) DILP release AA Sima (HIF-1a)

O2 Hph

Tor

Lipid mobilization Breathless Branchless (FGFR) (FGF) Trachea Fat body Brain

Fig. 8 HIF-1a/Sima is required in the fat body for hypoxia-induced growth suppression. a–c Fat-body-specific loss of sima via tissue-specific CRISPR/Cas9- mediated gene disruption reduces hypoxia-induced IPC DILP2 retention (a) and suppression of Dilp3 expression (b) and rescues hypoxia-induced growth restriction (c). Statistics: Student’s t-test for pairwise comparisons and one-way ANOVA with Dunnett’s for multiple-comparisons test. *P < 0.05, ***P < 0.001, compared with controls (ppl > Cas9/+ and UAS-simaKO/+ ). Error bars indicate SEM. Underlying data are provided in the Source Data file. a: n = 9. b: n = 6. c: n = 18–33. d Proposed model for integration of hypoxia and amino-acid sensing via the fat body and insulin regulation. Tracheation of the fat body is regulated by the FGF ligand Branchless secreted from this tissue in response to local tissue hypoxia, triggering tracheal outgrowth via the FGF receptor Breathless. When the tracheal system is unable to deliver sufficient oxygen because of environmental oxygen levels or body growth, Hph is unable to trigger the degradation of Sima/HIF-1a, which induces the expression or release of one or more humoral factors from the fat tissue that act on the insulin-producing cells (IPCs) to downregulate Dilp transcription and to block DILP release. Through a separate pathway, Hph inhibition also blocks Tor activity in the fat body, which alters lipid distribution. Hypoxia thereby reduces circulating insulin through activation of a central oxygen sensor in the fat tissue, which slows the growth of the whole organism. During these low-oxygen conditions, secretion of Branchless from the fat tissue promotes insulin-independent growth of the tracheal airways to increase oxygen supply. This system allows the organism to adapt its metabolism and growth to limited-oxygen conditions by reducing overall body growth through downregulation of insulin signaling, while at the same time promoting development of the tracheal system to maintain oxygen homeostasis

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y transcription was de-repressed (Fig. 8b) and the hypoxia-induced here is not. This observed transcriptional response could con- body-size reduction was significantly rescued by knockout of ceivably arise secondarily to DILP-release inhibition via autocrine sima in the fat body (Fig. 8c). Taken together, our data show that feedback regulation that operates in both Drosophila and fat-body HIF-1a/Sima activity is required for hypoxia-induced mammals60,61. However, lower secretion of DILPs, which we suppression of body growth via humoral signaling from the fat observe under hypoxic conditions, generally feeds back to that inhibits insulin secretion. induce an increase in the expression of Dilp3 and -5 rather than the decrease that we observe. Therefore, the hypoxia-induced alterations of Dilp3 and -5 expression appear to be specific Discussion transcriptional responses rather than feedback effects. Thus, In this report, we identify one tissue in particular, the fat body, beyond the identity of the fat-body factor involved, the which senses internal oxygen levels and regulates growth rate mechanisms operating in the IPCs by which it regulates insulin- accordingly. Our data show that, as an adaptive response to like gene expression and peptide release will be important to oxygen limitation, the fat tissue releases into the circulation one study in future experiments. or more factors that inhibit the secretion of insulin from the brain In mammals, several adipokines regulate β-cell function and to reduce systemic growth (Fig. 8d). We demonstrate that the insulin secretion, including Leptin, which conveys information ability of oxygen to reduce systemic body growth through about fat storage and is a functional analog of the Drosophila downregulation of insulin signaling requires Hph-dependent fat-derived cytokine Unpaired-29,62. Interestingly, we observed HIF-1a/Sima activity in the fat tissue. Furthermore, we show that strong increases in fat-body LD size induced by Tor inhibition hypoxia and AA deprivation both reduce Hph activity in the fat downstream of AA starvation, hypoxia, or Hph mutation, indi- tissue, and that this reduction leads to suppression of Tor sig- cating a change in lipid metabolism within the fat tissue. naling, independently of HIF-1a/Sima. This is consistent with a Although this phenotype is Tor-dependent and not upstream of requirement of Hph for cell growth49. In other contexts, Sima is the particular HIF-1a-dependent factor described above, it is known to regulate Tor-pathway activity via the protein Scylla/ possible that additional signals related to lipid metabolism may REDD158,59. However, this pathway does not appear to be be released by the fat body in response to hypoxia or starvation, responsible for the effects of hypoxia on Tor activity observed such as a lipid-binding protein or even a lipid per se. For example, here, as sima mutation does not block hypoxia- or starvation- the mammalian fatty acid-binding protein 4 is an insulin- induced Tor suppression. Likewise, Tor suppression is not modulating adipokine that is influenced by obesogenic conditions necessary for the systemic growth reduction induced by hypoxia. that lead to adipose tissue hypoxia63, and orthologous proteins Our data suggest that Hph is involved in AA sensing, in addition are encoded by the Drosophila genome. to its well-described role in oxygen sensing, and that HIF-1a is Most organisms stop growing after reaching a genetically not involved in this process. Together, this suggests that AA and predetermined species-characteristic size. Although insight from oxygen sensing converge through Hph in the fat body to mod- genetic studies in Drosophila1,6,12,16,17,55 into the mechanisms ulate systemic growth in response to environmental conditions. that regulate body growth with regard to nutrition helps to Many of the known Drosophila adipokines that affect insulin explain how organisms modulate their growth rate according secretion from the IPCs are regulated by Tor-pathway activity to nutritional conditions, a mechanism that allows organisms to in the fat body, including CCHa-210, Egr17, FIT11, GBP1 and assess their size and stop their growth when they have reached an GBP212, and Sun16. Our finding that AA availability regulates optimum has remained elusive. However, recent evidence sug- Hph activity, and that Hph activity modulates Tor signaling, thus gests that body size in insects may be determined by a mechanism places Hph upstream of these known factors, in addition to the that involves oxygen sensing3, and oxygen availability is known to separate Sima-dependent and Tor-independent humoral factor(s) place limits on insect body size18,20–22. According to this recent that modulate insulin secretion under hypoxia. Several routes insight, the limited growth ability of the tracheal system during by which AA availability regulates Tor have been investigated49 development may limit overall body size via downstream oxygen and Hph may modulate some of these and not others, thereby sensing3. The size of the tracheal system is established at the allowing for different responses to AA starvation and hypoxia. beginning of each developmental stage and remains largely fixed, Indeed, our work shows that HIF-1a/Sima is required for the aside from terminal branching, as the body grows until it even- growth-suppressive effect of hypoxia, but not for growth tually reaches the limit of the system’s ability to deliver oxygen. responses to varied dietary AA input, although Hph is involved in This allows the body to assess its size by sensing internal oxygen both. The mechanisms by which Hph activity, which simulta- concentrations and to terminate growth at a characteristic size neously requires AAs and oxygen, allows or promotes Tor sig- that is determined by the size of the tracheal system. Our RNAi naling is an interesting topic to investigate in future studies, screen shows that the FGF receptor Btl, which is a key factor as is the identity of the Tor-independent humoral factor(s) essential for tracheal growth during development, is a main downstream of Sima. determinant of body size—indeed, btl was a stronger hit than Our results show that hypoxia or loss of fat-body Hph activity known size-governing genes—and our data therefore support the Sima dependently represses Dilp3 and Dilp5 transcription, while notion that the tracheal system and oxygen sensing may be part having little or no suppressive effect on Dilp2 expression. This of a size-assessment mechanism. suggests that specific transcriptional regulation of Dilp3 and Dilp5 Oxygen homeostasis also requires the coordination of growth is an important component of the response to hypoxia. Con- between the tissues that consume oxygen and those that deliver it. sistent with this observation, previous studies have shown that the The development of the oxygen delivery system is therefore oxygen transcription of Dilp2, -3, and -5 are independently regulated7. sensitive in both mammals and Drosophila.Inmammals,local Nutrient deprivation reduces expression of Dilp3 and -5, while tissue hypoxia promotes angiogenesis via induction of many pro- having no effect on Dilp2 expression, similar to the effects of angiogenic factors, including FGF64.InDrosophila, tissue hypoxia exposure to hypoxia. This is consistent with our finding that both induces expression of the FGF-like ligand Bnl, leading to branching AA deprivation and hypoxia suppress Hph activity, although the of the tracheal airway tubes toward oxygen-deficient areas29,65. downstream pathways involved appear likely to be different, as Our study shows that this mechanism operates independently of at least some aspects of nutrient deprivation are relayed through insulin, as reduced insulin signaling in the trachea has no effect on the Tor pathway, whereas the hypoxia-specific signal(s) shown overall body growth. This system therefore allows an adaptive

12 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE response to low oxygen by reducing overall body growth via other components held constant. The AkhR-GAL4::p65 transgene containing the suppression of insulin signaling, while promoting hypoxia-induced enhanced activator GAL4::p6576 drives very strongly in the larval fat body. The fi FGF-dependent tracheal growth to increase oxygen delivery. GAL4::p65 was recombineered into genomic bacterial arti cial chromosome con- struct CH322-147J1777 obtained from Children’s Hospital Oakland Research Cell and tissue hypoxia are also observed in human conditions Institute (Oakland, CA, USA), replacing the first coding exon, and the construct of obesity and cancer. The insect fat body performs the functions was integrated into the attP2 genomic site by standard injection methods. The of mammalian fat and liver tissues. Accordingly, perturbation of following stocks were obtained from the Bloomington Drosophila stock center: arm-GAL4 (#1560), btl-GAL4 (#41803), Cg-GAL4 (#7011), da-GAL4 (#55850), systemic insulin signaling by adipose and hepatic tissue hypoxia 78 66 elav-GAL4 (#458), Mef2-GAL4 (#27390), ppl-GAL4 (#58768), R96A08-GAL4 is also observed in mammalian systems . In mammals, obesity (#48030), repo-GAL4 (#7415), simaKG07607 (sima null)56 (#14640), tGPH insulin- induces hypoxia within adipose tissue due to the rarefaction of activity sensor53 (#8164), UAS-Cas9.P2 (#58986), UAS-CCHa2-RNAi (#57183), vascularization of this tissue67, leading to the release of inflam- UAS-CCHa2-R-RNAi #2 (#25855), UAS-eiger-RNAi (#58993), UAS-InR-RNAi matory mediators and other adipokines that are associated with (#992), UAS-methuselah-RNAi #1 (#27495) and #2 (#36823), UAS-methuselah-like- 10-RNAi #1 (#62315) and #2 (#51753), UAS-sima-RNAi (#26207), UAS-TorTED the pathophysiology of obesity-related metabolic disorders (Tor dominant negative; #7013), and UAS-Tsc1/2 and UAS-VALIUM10-Luciferase 68–73 including diabetes . Although loss of normal β-cell activity (#35789). Vienna Drosophila Resource Center (VDRC) RNAi lines used for follow- is considered a main factor in diabetes, the mechanism by which up studies include UAS-Akt-RNAi (#103703), UAS-bnl-RNAi (#101377), UAS-btl- tissue hypoxia affects insulin secretion is poorly understood. RNAGD950 (#950), UAS-btl-RNAiGD27108 (#27108), UAS-CCHa2-R-RNAi #1 fi (#100290)10, UAS-grnd-RNAi #1 (#104538)17 and #2 (#43454)17, UAS-sima-RNAi Our nding of one or more hypoxia-induced fat-body-derived (#106187), UAS-daw-RNAi (#105309), and UAS-Stunted-RNAi (#23685)16. HphΔ9 insulinostatic factors may lead to insights into the role of adipose- (fatigaΔ9, fgaΔ9)56 was a kind gift of P. Wappner (Instituto Leloir), ubi-GFP::ODD tissue hypoxia in obesity and its impact on diabetes. Furthermore, hypoxia sensor (more rigorously, Hph-activity sensor)54 was a kind gift of S. we show a link between oxygen and AA availability in the adipose Luschnig (U. Münster), and UAS-FLAG::Dilp279 and UAS-RhebEP50.084,w-/TM6B, 80 51 tissue through Hph-dependent regulation of the Tor pathway, Tb were gifts of E. Hafen and H. Stocker (ETH Zürich). UAS-DILP2HF was a kind gift from S. Park and S. Kim (Stanford). The common laboratory stock w1118 linking these pathways in a common metabolic response to from VDRC (#60000) was used as a control in many experiments. oxygen limitation and nutrient scarcity. Obesity also causes physical and hormonal changes that affect RNAi screen for pupal size. A list of genes encoding the secretome (secreted breathing patterns, leading to apnea and thus intermittent epi- proteins) and receptome (membrane-associated proteins) was generated using 74 sodes of systemic hypoxia . These hypoxic periods can induce ENSEMBL and GLAD gene ontology databases. UAS-RNAi lines against 1845 changes in the liver, leading to fatty liver disease and dyslipide- genes for the screening of the secretome and receptome were obtained from the mia. The alterations to fat-body lipid metabolism observed here Bloomington Drosophila stock center32 and the VDRC31, and transformant IDs are given in Supplementary Data 1. Males from each RNAi line were crossed to may thus be relevant to human health as well. Furthermore, da-GAL4 virgin females, animals were allowed to lay eggs for 24 h, and the average hypoxia-induced programs play important roles in tumor for- size of pupal offspring was determined from images acquired with a Point Grey mation. During cancer development, tumor cells undergo a Grasshopper3 camera using a custom script in MATLAB (The MathWorks, Inc., metabolic reprogramming, the so-called Warburg effect, in which Natick, Massachusetts, USA)81. Each line was given a Z-score, calculated as the number of SD between that line’s average and the average of all lines. Lines with their metabolism shifts from oxidative phosphorylation to gly- − 75 a Z-score between 2 and 2, corresponding in this data set roughly to a ±15% colysis , and activation of HIF-1a is believed to play a key role in size abnormality, were not analyzed further, whereas 89 lines with greater effect this shift. As the hypoxia-sensing mechanism and the insulin- were identified. signaling system are conserved between flies and mammals, understanding the effects of hypoxia on the fat body could thus Generation of tissue-specific CRISPR lines. To generate lines for tissue-specific provide insight into many human disease states. It will be of CRISPR-based deletion of Hph and sima/HIF-1a, double guide RNAs were cloned interest to study whether tissue hypoxia also inhibits Tor-pathway into plasmid pCFD682,83 (Addgene #73915). The Hph construct should lead to the activity in mammalian adipocytes. deletion of part of the genomic region encoding the dioxygenase domain common to all Hph isoforms, and the sima gRNAs target the region encoding that protein’s In conclusion, our study unravels a mechanism that allows DNA-binding domain. The Cas9 target sequences below were synthesized as part organisms to adapt their metabolism and growth to environments of long oligonucleotides that were used to amplify gRNA structural sequences from with low oxygen. Hypoxia activates a fat-tissue oxygen sensor that pCFD6 using Q5 polymerase (New England Biolabs, #M0491S); plasmid template remotely controls the secretion of insulin from the brain by inter- DNA was then removed by DpnI digestion (NEB, #R0176S). pCFD6 was linearized by BbsI digestion (NEB, #R3539S) and treated with alkaline phosphatase (NEB, organ communication. This involves the inhibition of Hph #M0290S) to prevent self-ligation. The PCR products were recombined into the activity, leading to the activation of a HIF-1a-dependent genetic linearized, phosphatase-treated vector by Gibson assembly using the NEBuilder program within the fat tissue, which then secretes one or more HiFi DNA Assembly Master Mix (NEB, #E2621S). Clones were analyzed by PCR humoral signals that alter insulin-gene expression and repress and sequence confirmed. pCFD6-UAS-HphKO and pCFD6-UAS-simaKO were insulin secretion, thereby slowing growth. We also show that AA integrated into landing site attP2 (third chromosome) in-house and by BestGene fi (Chino Hills, CA) using standard methods. To remove fatiga or sima function scarcity, like oxygen de ciency, inhibits Hph activity, and that the in the fat body, two CRISPR-driver stocks—ppl-GAL4; UAS-Cas9.P2/TM6B, Tb activity of Hph, but not of HIF-1a, is required for Tor activity in and Cg-GAL4; UAS-Cas9.P2/TM6B, Tb—were made, and these were crossed to the the fat body. Thus, in addition to its role in regulating the as-yet UAS-gRNA lines generated above, or to w1118 as a driver control. The Cas9.P2 unidentified fat-body hypoxia signal via HIF-1a, Hph connects variant used here is somewhat weak, so to increase Cas9 expression by increasing GAL4 activity, we incubated the offspring at 30 °C. Hph target sites: 5′- both oxygen and AA levels to the Tor pathway through an GTATCGTGTTAACACGATGA-3′ and 5′-GAATATCAACTGGGATGCGC-3′; unknown HIF-1a-independent mechanism. Given the conserva- sima target sites: 5′-GGCTAGATGTCGTCGCTCCA-3′ and 5′-GGCTA- tion of oxygen-sensing and growth-regulatory systems, and the GATGTCGTCGCTCCA-3′. influence of oxygen on growth between Drosophila and mam- mals, a similar adaption response may operate in mammals via Tracheal measurements and hypoxic survival rates. To determine tracheal adipose tissue oxygen sensing to maintain homeostasis. length and branching, L3 larvae [96 h after egg lay (AEL) for control; 120 h for btl- RNAi animals to account for slowed growth] were heat-fixed in a drop of glycerol on a microscope slide on a 60 °C heat block until they became relaxed. Tracheal Methods cells and branching of the fat-body terminal branches were visualized in whole Fly strains and husbandry. Unless specified otherwise, animals were reared on larvae by imaging of tracheae labeled with GFP or by direct light microscopy. standard cornmeal-yeast medium (Nutri-Fly, Bloomington recipe) at 25 °C and Length and branching were manually quantified using the FIJI (NIH) software 60% relative humidity, under a 12L:12D light cycle. The 1× yeast diet contains 34 g package. To determine survival rates of btl knockdown animals, timed egg lays baker’s yeast, 6 g agar, 60 g sucrose, 84 g cornmeal, 4.8 mL propionic acid, and 1.6 g were performed, and first-instar (L1) larvae were transferred to fly vials containing Tegosept/Nipagin antifungal agent (in 16 mL 100% ethanol) per liter. Restric- standard food and incubated in either hypoxia (5% O2) or normoxia (21% O2, ted (0.1× and 0.3×) diets contain 3.4 g/L or 10 g/L yeast, respectively, with all other normal atmospheric oxygen concentration), and the number of animals surviving components held constant. Sugar-only diet contains no cornmeal or yeast with to the pupal stage was determined. The low-oxygen environment was generated by

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

+ slowly passing 5% O2 95% N2 through a clear, airtight plastic chamber placed Immunostaining. Brains were removed from ex-vivo cultures or dissected from inside a 25 °C, 12L:12D incubator. Gas was bubbled through water upstream and larvae 96 h AEL (normoxia) or 120 h AEL (hypoxia and Hph mutants) in cold downstream of the chamber, to provide humidity and to prevent infiltration of Schneider’s culture medium and fixed in fresh 4% paraformaldehyde in PBS for outside air, and vented outside the incubator. The chamber was flushed with high- 1 h. Tissues were washed four times in PBST (PBS containing 0.1% Triton X-100), flow-rate gas for 5 min after any opening. blocked with PBST containing 3% normal goat serum (Sigma-Aldrich #G9023) at room temperature for 1 h, and incubated with primary antibodies overnight at 4 °C. The following primary antibodies were used: rat anti-DILP252 (kind gift of Analysis of developmental timing and growth rates. To synchronize larvae for P. Léopold, U. Nice) at 1:250 dilution; mouse anti-DILP3 at 1:200 dilution (kind growth rate and developmental-timing assays, flies were allowed to lay eggs for 4 h gift of J. Veenstra, U. Bordeaux); rabbit anti-DILP584 at 1:2,000 dilution (kind gift on apple-juice agar plates supplemented with yeast paste. Newly hatched L1 larvae of D. Nässel, U. Stockholm); rabbit anti-phospho-ribosomal pS6 (anti-pS6)57 at were transferred into vials with standard food (30 per vial to prevent crowding) 1:200 (kindly given by A. Teleman, DKFZ). Tissues were washed three times in 24 h later. To determine growth rates, synchronized larvae were weighed at time PBST and incubated in secondary antibodies: goat anti-rabbit Alexa Fluor 488 points 75–96 h AEL in groups of 5–10 animals. For developmental-timing (Thermo Fisher Scientific, #A32731), goat anti-rat Alexa Fluor 555 (Thermo Fisher experiments, pupariation time was determined manually. Scientific, #A21434), and goat anti-mouse Alexa Fluor 647 (Thermo Fisher Sci- entific, #A28181), all diluted 1:500 in PBST, overnight at 4 °C. After four washes in PBST, tissues were mounted in Vectashield (Vector Labs #H-1000) and imaged Ex-vivo organ co-culture. For organ co-culture experiments, brains and fat using a Zeiss LSM 800 confocal microscope and Zen software. Retained insulin 1118 bodies were dissected from 96 h AEL wild-type w larvae or 120 h AEL Hph- levels were calculated using the FIJI software package. mutant larvae (to adjust for their slow growth) in cold Schneider’s culture medium (Sigma-Aldrich #S0146) containing 5% fetal bovine serum (Sigma). Fifty brains, μ with or without fat bodies present, were incubated in 50 L culture medium for Fat-body imaging. To measure in-vivo intracellular insulin-signaling transduction 16 h at 25 °C in small dishes in moist chambers in either hypoxia or normoxia. in the fat body, we made use of a GFP-Pleckstrin Homology domain fusion protein After incubation, brains were removed for immunostaining and quantitative real- (GPH) expressed from the ubiquitously active β-Tubulin promoter (tGPH). time PCR (qPCR) analyses. Insulin-signaling activity causes recruitment of the GPH protein to the plasma membrane53. Hypoxia was induced at 72 h AEL, and hypoxic animals were assayed for fat-body GFP localization 24 h later together with age-matched 96 h AEL Expression analysis by qPCR. To quantify gene expression by qPCR, mRNA was normoxic controls. Fat bodies were dissected in cold Schneider’s medium and prepared from samples of five whole larvae each (96 h AEL for btl > + controls and immediately imaged. To determine whether the fat body experiences biologically normoxia animals, and 120 h AEL for btl-RNAi larvae, animals raised under relevant hypoxia under 5% ambient oxygen levels and whether Hph activity is hypoxia, and Hph mutants to adjust for their slower growth) using the RNeasy responsive to AA levels, we used a GFP::ODD sensor line, made up of the ODD Mini Kit (Qiagen #74106) with DNase treatment (Qiagen #79254). RNA yield was of Sima fused to GFP, expressed throughout the animal from the ubiquitin-69E determined using a NanoDrop spectrophotometer (ThermoFisher Scientific), and promoter54. Similar to endogenous Sima protein, the GFP::ODD reporter is cDNA was generated using the High-Capacity cDNA Reverse Transcription Kit degraded at a rate reflecting Hph activity; thus, in normal conditions, the GFP with RNase Inhibitor (ThermoFisher #4368814). qPCR was performed using the signal is low, whereas in hypoxia or other conditions of Hph inhibition, the GFP QuantiTect SYBR Green PCR Kit (Fisher Scientific #204145) and an Mx3005P signal is increased. Unmodified RFP expressed from the same regulatory elements qPCR System (Agilent Technologies). Expression levels were normalized against provides a ratiometric denominator. For hypoxia tests, larvae were reared under RpL32. FlyPrimerBank pre-computed oligo pairs were used for some target genes. normoxia or hypoxia until 96 or 120 h AEL, respectively, at which point fat bodies The primer pairs used are given in Supplementary Table 1. were dissected in cold Schneider’s medium and immediately imaged. For protein- starvation assays, animals were reared in normoxia at 25 °C on standard (1× yeast) Western blotting analysis. To quantify signaling pathway activity downstream of diets until 92 h AEL and then transferred either to a fresh vial of the same diet or to InR, the level of Akt phosphorylation (pAkt) was determined by western blotting. protein-starvation medium (sugar alone) for 4 h; fat bodies were dissected in cold Schneider’s medium (non-starved animals) or cold PBS (starved animals, to pre- For hypoxia experiments, larvae were transferred to 5% O2 at 72 h AEL and fl analyzed 24 h later, compared with 96 h AEL normoxic controls. Hph-mutant vent uptake of AAs). Confocal stacks of endogenous xFP uorescence were taken larvae were assayed at 120 h AEL. For each sample, three larvae were homogenized immediately. Levels of GFP and RFP signals were integrated using the FIJI package, in SDS loading buffer (Bio-Rad), boiled for 5 min, centrifuged at 14,000 × g for and the ratio between them was calculated. LDs were imaged using label-free – Coherent Anti-Stokes Raman Scattering (CARS) microscopy. For normoxia assays, 5 min to pellet debris, and electrophoresed through a precast 4 20% poly- Δ9 acrylamide gradient gel (Bio-Rad). Proteins were transferred to a polyvinylidene animals were reared on standard diet in normoxia until harvest (for Hph ,at fl 120 h; otherwise, 96 h). For hypoxia assays, w1118 and simaKG07607 animals were di uoride membrane (Millipore), and the membrane was blocked with Odyssey Δ9 Blocking Buffer (LI-COR) and incubated with rabbit anti-pAkt (Cell Signaling reared for 120 h in hypoxia; Hph animals died when reared in hypoxia, so these Technology #4054, diluted 1:1,000) or rabbit anti-pan-Akt (Cell Signaling Tech- animals were reared in normoxia for 120 h and then transferred to hypoxia for 16 h α before harvest. For measurements of response to AA starvation, w1118 and nology #4691, diluted 1:1,000), and mouse -Tubulin (Sigma #T9026, diluted KG07607 Δ9 1:5,000) in Odyssey Blocking Buffer (LI-COR) containing 0.2% Tween 20. Primary sima animals were reared on standard (1× yeast) for 96 h and Hph for — 120 h, and then transferred to sugar-only diet for 16 h. Fat bodies were dissected in antibodies were detected with goat secondary antibodies IRDye 680RD anti- ’ mouse (LI-COR #925-68070) and IRDye 800CW anti-rabbit (LI-COR #925-32210) cold Schneider s culture medium and immediately imaged using a Leica TCS SP8 — confocal microscope85,86, using a picoEmerald laser and BS635 filter for lipids diluted 1:10,000 and bands were visualized using an Odyssey Fc imaging system – (LI-COR). Uncropped blots are presented in the Source Data file. along with these settings: IR power, 2000 mW; OPO, 2000 mW at 780 940 nm, pulse width 5–6 ps, 80 MHz; pump, 1174 mW at 1064 nm, pulse length 7 ps, 80 MHz. To image lipids, the lasers were tuned to the C–H stretching vibration: ELISA of circulating tagged DILP2. Dilp2 > DILP2HF51 animals were reared pump beam, 816.4 nm; Stokes beam, 1,064.6 nm; laser output, 1.3 W; scan speed, under normal conditions until 24 h before wandering and then transferred to 400 Hz. Signals were collected from the epi-CARS and epi-SHG detectors. LD size starvation medium (1% agar in water) in normoxia, to normal food under hypoxia was quantified using the FIJI software package. (5% O2), or back to normoxia and normal diet for 12 h. Hemolymph was extracted by cutting the larval cuticle and pipetting the released droplet into a collection vial. Collected hemolymph was heat-treated at 60 °C for 5 min, cen- RNAi mini-screen of fat-body factors and IPC receptors. R96A08-GAL4, tar- trifuged to remove any debris or aggregates, and used in an ELISA against the geting the IPCs, was crossed to RNAi constructs against receptors, and the fat-body HA and FLAG tags. F8 MaxiSorp Nunc-Immuno modules (Thermo Scientific driver AkhR-GAL4::p65 was crossed to lines targeting secreted factors. Fat-body #468667) were incubated at 4 °C overnight with 5 μg/mL anti-FLAG (Sigma- driver ppl-GAL4 was crossed to daw-RNAi and UAS-Rheb. Newly hatched larvae Aldrich #F1804) in 200 mM NaHCO3 buffer (pH 9.4). After two washes with from 4 h egg lays were transferred at 24 h AEL to standard food vials and incubated phosphate-buffered saline (PBS) + 0.1% Triton X-100 (PBST), the plate was at 25 °C in normoxia or hypoxia (5% O2) until pupariation. Pupal size was mea- blocked with PBST + 4% non-fat dry milk for 2 h at room temperature and sured, and for each genotype the ratio for each possible hypoxic and normoxic vial washed again three times in PBST. One microliter of hemolymph or synthetic pair was computed to make up each data set. Experimental data were compared HA::spacer::FLAG peptide standard (DYKDDDDKGGGGSYPYDVPDYamide) with data from controls expressing UAS-Luciferase in a background matching that was diluted into 50 μLPBST+ 25 ng/mL mouse anti-HA peroxidase (Roche of the RNAi lines, except for the ΔGbp1,2 deletion line, which was normalized to 12013819001) + 1% non-fat dry milk, added to the wells, and incubated overnight w1118, and ppl> crosses, which were compared with ppl-GAL4/+. at 4 °C. Fluid was removed and wells were washed in PBST six times. One-step Ultra TMB ELISA substrate (Thermo Scientific #34028) was added to each well (100 μL) and incubated for 15 min at room temperature. Sulfuric acid (2 M, 100 Statistical analyses. Statistical tests and P-values are given within each figure μL) was added to stop the reaction and absorbance at 450 nm was measured using legend. Statistics were calculated using GraphPad Prism Software, applying analysis a plate reader. Plate, peptide standard, UAS-DILP2HF, anti-FLAG, anti-HA, of variance, Dunnett’s test, or Kruskal–Wallis non-parametric tests for multiple- substrate, and detailed protocol were all generously provided by S. Park and comparisons tests and two-tailed Student’s t-tests or two-tailed exact S. Kim (Stanford University). Mann–Whitney non-parametric tests for pairwise comparisons. Error bars indicate

14 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y ARTICLE

SEM and P-values are indicated as follows: *P < 0.05, **P < 0.01, ***P < 0.001, and 22. Frazier, M. R., Woods, H. A. & Harrison, J. F. Interactive effects of rearing ****P < 0.0001. temperature and oxygen on the development of Drosophila melanogaster. Physiol. Biochem. Zool. 74, 641–650 (2001). Reporting summary. Further information on experimental design is available in 23. Frisancho, A. R. Developmental functional adaptation to high altitude: review. – the Nature Research Reporting Summary linked to this article. Am. J. Hum. Biol. 25, 151 168 (2013). 24. Frisancho, A. R. & Baker, P. T. Altitude and growth: a study of the patterns of physical growth of a high altitude Peruvian Quechua population. Am. J. Data availability Phys. Anthr. 32, 279–292 (1970). Data that support the findings of the current study are available from the corresponding 25. Van Voorhies, W. A. Metabolic function in Drosophila melanogaster in author on reasonable request. response to hypoxia and pure oxygen. J. Exp. Biol. 212, 3132–3141 (2009). 26. Harrison, J. F., Shingleton, A. W. & Callier, V. Stunted by developing in Received: 15 June 2018 Accepted: 10 April 2019 hypoxia: linking comparative and model organism studies. Physiol. Biochem Zool. 88, 455–470 (2015). 27. Gorr, T. A., Gassmann, M. & Wappner, P. Sensing and responding to hypoxia via HIF in model invertebrates. J. Insect Physiol. 52, 349–364 (2006). 28. Klambt, C., Glazer, L. & Shilo, B. Z. breathless, a Drosophila FGF receptor homolog, is essential for migration of tracheal and specific midline glial cells. References Genes Dev. 6, 1668–1678 (1992). 1. Danielsen, E. T., Moeller, M. E. & Rewitz, K. F. Nutrient signaling and 29. Jarecki, J., Johnson, E. & Krasnow, M. A. Oxygen regulation of airway – developmental timing of maturation. Curr. Top. Dev. Biol. 105,37–67 (2013). branching in Drosophila is mediated by branchless FGF. Cell 99, 211 220 2. Tennessen, J. M. & Thummel, C. S. Coordinating growth and maturation - (1999). insights from Drosophila. Curr. Biol. 21, R750–R757 (2011). 30. Sutherland, D., Samakovlis, C. & Krasnow, M. A. branchless encodes a 3. Callier, V. & Nijhout, H. F. Control of body size by oxygen supply reveals size- Drosophila FGF homolog that controls tracheal cell migration and the pattern – dependent and size-independent mechanisms of molting and metamorphosis. of branching. Cell 87, 1091 1101 (1996). Proc. Natl Acad. Sci. USA 108, 14664–14669 (2011). 31. Dietzl, G. et al. A genome-wide transgenic RNAi library for conditional gene – 4. Callier, V. et al. The role of reduced oxygen in the developmental physiology inactivation in Drosophila. Nature 448, 151 156 (2007). of growth and metamorphosis initiation in Drosophila melanogaster. 32. Ni, J. Q. et al. A genome-scale shRNA resource for transgenic RNAi in – J. Exp. Biol. 216, 4334–4340 (2013). Drosophila. Nat. Methods 8, 405 407 (2011). 5. Hietakangas, V. & Cohen, S. M. Regulation of tissue growth through nutrient 33. Chen, C., Jack, J. & Garofalo, R. S. The Drosophila insulin receptor is required – sensing. Annu Rev. Genet. 43, 389–410 (2009). for normal growth. Endocrinology 137, 846 856 (1996). fi 6. Boulan, L., Milan, M. & Leopold, P. The systemic control of growth. Cold 34. Loren, C. E. et al. Identi cation and characterization of DAlk: a novel Spring Harb. Perspect. Biol. 7, pii: a019117 https://doi.org/10.1101/ Drosophila melanogaster RTK which drives ERK activation in vivo. Genes Cells – cshperspect.a019117 (2015). 6, 531 544 (2001). 7. Ikeya, T., Galic, M., Belawat, P., Nairz, K. & Hafen, E. Nutrient-dependent 35. Okamoto, N. & Nishimura, T. Signaling from glia and cholinergic neurons expression of insulin-like peptides from neuroendocrine cells in the CNS controls nutrient-dependent production of an insulin-like peptide for – contributes to growth regulation in Drosophila. Curr. Biol. 12, 1293–1300 Drosophila body growth. Dev. Cell 35, 295 310 (2015). (2002). 36. Diaz-Benjumea, F. J. & Garcia-Bellido, A. Behaviour of cells mutant for an 8. Wang, S., Tulina, N., Carlin, D. L. & Rulifson, E. J. The origin of islet-like cells EGF receptor homologue of Drosophila in genetic mosaics. Proc. Biol. Sci. 242, – in Drosophila identifies parallels to the vertebrate endocrine axis. Proc. Natl 36 44 (1990). ’ Acad. Sci. USA 104, 19873–19878 (2007). 37. Rewitz, K. F., Yamanaka, N., Gilbert, L. I. & O Connor, M. B. The insect 9. Rajan, A. & Perrimon, N. Drosophila cytokine unpaired 2 regulates neuropeptide PTTH activates receptor tyrosine kinase torso to initiate – physiological homeostasis by remotely controlling insulin secretion. Cell 151, metamorphosis. Science 326, 1403 1405 (2009). 123–137 (2012). 38. Danielsen, E. T. et al. Transcriptional control of steroid biosynthesis genes in 10. Sano, H. et al. The nutrient-responsive hormone CCHamide-2 controls the Drosophila prothoracic gland by ventral veins lacking and knirps. PLoS growth by regulating insulin-like peptides in the brain of Drosophila Genet. 10, e1004343 (2014). melanogaster. PLoS Genet 11, e1005209 (2015). 39. Bialecki, M., Shilton, A., Fichtenberg, C., Segraves, W. A. & Thummel, C. S. 11. Sun, J. et al. Drosophila FIT is a protein-specific satiety hormone essential for Loss of the ecdysteroid-inducible E75A orphan nuclear receptor uncouples – feeding control. Nat. Commun. 8, 14161 (2017). molting from metamorphosis in Drosophila. Dev. Cell 3, 209 220 (2002). 12. Koyama, T. & Mirth, C. K. Growth-blocking peptides as nutrition-sensitive 40. Huang, X., Warren, J. T., Buchanan, J., Gilbert, L. I. & Scott, M. P. Drosophila signals for insulin secretion and body size regulation. PLoS Biol. 14, e1002392 Niemann-Pick type C-2 genes control sterol homeostasis and steroid (2016). biosynthesis: a model of human neurodegenerative disease. Development 134, – 13. Meschi, E., Leopold, P. & Delanoue, R. An EGF-responsive neural circuit 3733 3742 (2007). couples insulin secretion with nutrition in Drosophila. Dev. Cell 48,76–86 e75 41. Wilk, R., Weizman, I. & Shilo, B. Z. trachealess encodes a bHLH-PAS protein – (2019). that is an inducer of tracheal cell fates in Drosophila. Genes Dev. 10,93 102 14. Ghosh, A. C. & O’Connor, M. B. Systemic Activin signaling independently (1996). regulates sugar homeostasis, cellular metabolism, and pH balance in 42. Ohshiro, T., Emori, Y. & Saigo, K. Ligand-dependent activation of breathless – Drosophila melanogaster. Proc. Natl Acad. Sci. USA 111, 5729–5734 (2014). FGF receptor gene in Drosophila developing trachea. Mech. Dev. 114,3 11 15. Bai, H., Kang, P., Hernandez, A. M. & Tatar, M. Activin signaling targeted by (2002). insulin/dFOXO regulates aging and muscle proteostasis in Drosophila. PLoS 43. Glazer, L. & Shilo, B. Z. The Drosophila FGF-R homolog is expressed in the Genet. 9, e1003941 (2013). embryonic tracheal system and appears to be required for directed tracheal – 16. Delanoue, R. et al. Drosophila insulin release is triggered by adipose Stunted cell extension. Genes Dev. 5, 697 705 (1991). ligand to brain Methuselah receptor. Science 353, 1553–1556 (2016). 44. Kim, J. & Neufeld, T. P. Dietary sugar promotes systemic TOR activation 17. Agrawal, N. et al. The Drosophila TNF Eiger is an adipokine that acts on in Drosophila through AKH-dependent selective secretion of Dilp3. Nat. insulin-producing cells to mediate nutrient response. Cell Metab. 23, 675–684 Commun. 6, 6846 (2015). (2016). 45. Puig, O. & Tjian, R. Transcriptional feedback control of insulin receptor by – 18. Palos, L. A. & Blasko, G. Effect of hypoxia on the development of Drosophila dFOXO/FOXO1. Genes Dev. 19, 2435 2446 (2005). melanogaster (Meigen). Aviat. Space Environ. Med. 50, 411–412 (1979). 46. Colombani, J. et al. Antagonistic actions of ecdysone and insulins determine fi – 19. Callier, V. & Nijhout, H. F. Body size determination in insects: a review and nal size in Drosophila. Science 310, 667 670 (2005). synthesis of size- and brain-dependent and independent mechanisms. Biol. 47. Jones, T. A. & Metzstein, M. M. Examination of Drosophila larval tracheal Rev. Camb. Philos. Soc. 88, 944–954 (2013). terminal cells by light microscopy. J. Vis. Exp., e50496, https://doi.org/ 20. Henry, J. R. & Harrison, J. F. Plastic and evolved responses of larval tracheae 10.3791/50496 (2013). and mass to varying atmospheric oxygen content in Drosophila melanogaster. 48. Nambu, J. R., Chen, W., Hu, S. & Crews, S. T. The Drosophila melanogaster J. Exp. Biol. 207, 3559–3567 (2004). similar bHLH-PAS gene encodes a protein related to human hypoxia- – 21. Peck, L. S. & Maddrell, S. H. P. Limitation of size by hypoxia in the fruit fly inducible factor 1 alpha and Drosophila single-minded. Gene 172, 249 254 Drosophila melanogaster. J. Exp. Zool. A Comp. Exp. Biol. 303A, 968–975 (1996). (2005). 49. Frei, C. & Edgar, B. A. Drosophila cyclin D/Cdk4 requires Hif-1 prolyl hydroxylase to drive cell growth. Dev. Cell 6, 241–251 (2004).

NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09943-y

50. Bruick, R. K. & McKnight, S. L. A conserved family of prolyl-4-hydroxylases 79. Honegger, B. et al. Imp-L2, a putative homolog of vertebrate IGF-binding that modify HIF. Science 294, 1337–1340 (2001). protein 7, counteracts insulin signaling in Drosophila and is essential for 51. Park, S. et al. A genetic strategy to measure circulating Drosophila insulin starvation resistance. J. Biol. 7, 10 (2008). reveals genes regulating insulin production and secretion. PLoS Genet. 10, 80. Stocker, H. et al. Rheb is an essential regulator of S6K in controlling cell e1004555 (2014). growth in Drosophila. Nat. Cell Biol. 5, 559–565 (2003). 52. Geminard, C., Rulifson, E. J. & Leopold, P. Remote control of insulin secretion 81. Moeller, M. E. et al. Warts signaling controls organ and body growth through by fat cells in Drosophila. Cell Metab. 10, 199–207 (2009). regulation of ecdysone. Curr. Biol. 27, 1652–1659 e1654 (2017). 53. Britton, J. S., Lockwood, W. K., Li, L., Cohen, S. M. & Edgar, B. A. Drosophila’s 82. Port, F. & Bullock, S. L. Augmenting CRISPR applications in Drosophila with insulin/PI3-kinase pathway coordinates cellular metabolism with nutritional tRNA-flanked sgRNAs. Nat. Methods 13, 852–854 (2016). conditions. Dev. Cell 2, 239–249 (2002). 83. Huynh, N., Zeng, J., Liu, W. & King-Jones, K. A Drosophila CRISPR/Cas9 54. Misra, T. et al. A genetically encoded biosensor for visualising hypoxia toolkit for conditionally manipulating gene expression in the prothoracic responses in vivo. Biol. Open 6, 296–304 (2017). gland as a test case for polytene tissues. G3 8, 3593–3605 (2018). 55. Colombani, J. et al. A nutrient sensor mechanism controls Drosophila growth. 84. Soderberg, J. A., Birse, R. T. & Nassel, D. R. Insulin production and signaling Cell 114, 739–749 (2003). in renal tubules of Drosophila is under control of tachykinin-related peptide 56. Centanin, L., Ratcliffe, P. J. & Wappner, P. Reversion of lethality and growth and regulates stress resistance. PLoS ONE 6, e19866 (2011). defects in Fatiga oxygen-sensor mutant flies by loss of hypoxia-inducible 85. Danielsen, E. T. et al. A Drosophila genome-wide screen identifies regulators factor-alpha/Sima. EMBO Rep. 6, 1070–1075 (2005). of steroid hormone production and developmental timing. Dev. Cell 37, 57. Romero-Pozuelo, J., Demetriades, C., Schroeder, P. & Teleman, A. A. CycD/ 558–570 (2016). Cdk4 and discontinuities in Dpp signaling activate TORC1 in the Drosophila 86. Hentze, J. L., Carlsson, M. A., Kondo, S., Nassel, D. R. & Rewitz, K. F. The wing disc. Dev. Cell 42, 376–387 e375 (2017). neuropeptide allatostatin A regulates metabolism and feeding decisions in 58. Reiling, J. H. & Hafen, E. The hypoxia-induced paralogs Scylla and Charybdis Drosophila. Sci. Rep. 5, 11680 (2015). inhibit growth by down-regulating S6K activity upstream of TSC in Drosophila. Genes Dev. 18, 2879–2892 (2004). 59. Brugarolas, J. et al. Regulation of mTOR function in response to hypoxia by Acknowledgements REDD1 and the TSC1/TSC2 tumor suppressor complex. Genes Dev. 18, We thank our Erasmus-program visiting undergraduate Niyati Jain for help. We are 2893–2904 (2004). grateful to Seung Kim, Pierre Léopold, Stefan Luschnig, Dick Nässel, Sangbin Park, 60. Gronke, S., Clarke, D. F., Broughton, S., Andrews, T. D. & Partridge, L. Aurelio Teleman, Jan Veenstra, and Pablo Wappner for kindly sharing antibodies, fly Molecular evolution and functional characterization of Drosophila insulin-like lines, and reagents. We also thank Addgene, the Bloomington Drosophila Stock Center, peptides. PLoS Genet 6, e1000857 (2010). and the Vienna Drosophila Resource Center for making plasmids and fly lines broadly 61. Xu, G. G. & Rothenberg, P. L. Insulin receptor signaling in the beta-cell available. This work was supported by the Novo Nordisk Foundation STAR grant and an influences insulin gene expression and insulin content: evidence for autocrine Innovation foundation grant to J.L.H. and K.F.R. and by Novo Nordisk Foundation grant beta-cell regulation. Diabetes 47, 1243–1252 (1998). 16OC0021270 and Danish Council for Independent Research Natural Sciences grant 62. Cantley, J. The control of insulin secretion by adipokines: current evidence for 4181-00270 to K.F.R. K.A.H. was supported by Villum Foundation grant 15365 to K.A.H. adipocyte-beta cell endocrine signalling in metabolic homeostasis. Mamm. Genome 25, 442–454 (2014). 63. Wu, L. E. et al. Identification of fatty acid binding protein 4 as an Author contributions adipokine that regulates insulin secretion during obesity. Mol. Metab. 3, M.J.T., A.F.J., C.F.C., T.K., A.M., D.K.S., D.F.M.M., E.T.D., S.K.P., K.A.H. and K.F.R. 465–473 (2014). performed the experiments and analyzed the data. M.J.T., A.F.J., C.F.C., J.L.H., K.A.H. 64. Krock, B. L., Skuli, N. & Simon, M. C. Hypoxia-induced angiogenesis: good and K.F.R. designed the experiments. M.J.T., K.A.H. and K.F.R. wrote the manuscript. and evil. Genes Cancer 2, 1117–1133 (2011). 65. Fong, G. H. Mechanisms of adaptive angiogenesis to tissue hypoxia. Angiogenesis 11, 121–140 (2008). Additional information 66. Regazzetti, C. et al. Hypoxia decreases insulin signaling pathways in Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- adipocytes. Diabetes 58,95–103 (2009). 019-09943-y. 67. Pasarica, M. et al. Reduced adipose tissue oxygenation in human obesity: evidence for rarefaction, macrophage chemotaxis, and inflammation without Competing interests: J.L.H. is an employee and shareholder of Novo Nordisk A/S. A.F.J. an angiogenic response. Diabetes 58, 718–725 (2009). is an employee of Novo Nordisk A/S. All remaining authors declare no competing 68. Polotsky, V. Y. et al. Intermittent hypoxia increases insulin resistance in interests. genetically obese mice. J. Physiol. 552, 253–264 (2003). 69. Norouzirad, R., Gonzalez-Muniesa, P. & Ghasemi, A. Hypoxia in obesity and Reprints and permission information is available online at http://npg.nature.com/ diabetes: potential therapeutic effects of hyperoxia and nitrate. Oxid. Med Cell reprintsandpermissions/ Longev. 2017, 5350267 (2017). 70. Rasouli, N. Adipose tissue hypoxia and insulin resistance. J. Invest. Med 64, Journal peer review information: Nature Communications thanks Dan Zhou and other 830–832 (2016). anonymous reviewer(s) for their contribution to the peer review of this work. Peer 71. Hosogai, N. et al. Adipose tissue hypoxia in obesity and its impact on reviewer reports are available. adipocytokine dysregulation. Diabetes 56, 901–911 (2007). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in 72. Lefere, S. et al. Hypoxia-regulated mechanisms in the pathogenesis of fi obesity and non-alcoholic fatty liver disease. Cell Mol. Life Sci. 73, 3419–3431 published maps and institutional af liations. (2016). 73. Kihira, Y. et al. Deletion of hypoxia-inducible factor-1alpha in adipocytes enhances glucagon-like peptide-1 secretion and reduces adipose tissue Open Access This article is licensed under a Creative Commons inflammation. PLoS ONE 9, e93856 (2014). Attribution 4.0 International License, which permits use, sharing, 74. Ryan, S. Adipose tissue inflammation by intermittent hypoxia: mechanistic adaptation, distribution and reproduction in any medium or format, as long as you give link between obstructive sleep apnoea and metabolic dysfunction. J. Physiol. appropriate credit to the original author(s) and the source, provide a link to the Creative 595, 2423–2430 (2017). Commons license, and indicate if changes were made. The images or other third party 75. Masson, N. & Ratcliffe, P. J. Hypoxia signaling pathways in cancer material in this article are included in the article’s Creative Commons license, unless metabolism: the importance of co-selecting interconnected physiological indicated otherwise in a credit line to the material. If material is not included in the pathways. Cancer Metab. 2, 3 (2014). article’s Creative Commons license and your intended use is not permitted by statutory fi 76. Pfeiffer, B. D. et al. Re nement of tools for targeted gene expression in regulation or exceeds the permitted use, you will need to obtain permission directly from – Drosophila. Genetics 186, 735 755 (2010). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 77. Venken, K. J. et al. Versatile P[acman] BAC libraries for transgenesis studies licenses/by/4.0/. in Drosophila melanogaster. Nat. Methods 6, 431–434 (2009). 78. Pfeiffer, B. D. et al. Tools for neuroanatomy and neurogenetics in Drosophila. Proc. Natl Acad. Sci. USA 105, 9715–9720 (2008). © The Author(s) 2019

16 NATURE COMMUNICATIONS | (2019) 10:1955 | https://doi.org/10.1038/s41467-019-09943-y | www.nature.com/naturecommunications